Give your own AI agent the whole QA job.
AxonQA speaks the Model Context Protocol, so Claude Code, Cursor, VS Code or an agent you wrote yourself can drive it directly. Not a handful of read-only endpoints: 200+ tools spanning the QA workspace from exploring your app to preparing a bug for review. Publicly reachable targets are explored in AxonQA Cloud by default; private or firewalled targets can use the local AxonQA Agent instead.
“Our checkout changed. Explore it, update the tests, run them, and tell me what broke.”
- 1
start_crawlcheckout journey9 pages, 34 elements - 2
generate_tests_from_crawlgrounded in what it found6 cases updated - 3
run_testsChrome and Firefoxrunning - 4
wait_for_run_completion5 passed, 1 failed - 5
get_failure_detailsdiscount fieldassertion mismatch, evidence collected - 6
raise_bugwith the recording attacheddraft ready for review
Checkout needs review. Six tests now cover the new flow. The discount assertion still fails, and a bug draft with the recording is ready for review. The other checks passed.
What your agent can reach
200+ tools, grouped by the job they do
The same core tool surface used by Axon AI is available to your agent. The live list is always at /api/mcp.
Explore the app
Point an agent at your application and let it map pages, elements, journeys and real locators before it writes anything.
start_crawlget_discovered_capabilitiescapture_domget_page_healthget_locator_libraryand moreCreate tests
Turn requirements, or what the exploration found, into structured test cases. Describe one test and get exactly that test.
describe_testgenerate_tests_from_crawlgenerate_testsimprove_testsexport_testsand moreSet up the workspace
Create the structure around the tests: projects, epics, stories, and test plans with runs and lifecycle states.
create_projectcreate_epiccreate_storycreate_test_planrun_test_planand moreAutomate and run
Generate a maintainable framework, edit its code, run it across browsers, and rerun only what failed.
generate_automation_codeedit_framework_filerun_testsrerun_failed_testsrun_test_chainand moreHeal and diagnose
Repair locator drift without changing assertions, inspect failure evidence, and reverse any suggested fix you disagree with.
get_heal_suggestionsapply_healsrevert_healsdebug_test_failurerun_visual_comparisonand moreBuild API suites
Collections, folders, endpoints, chained flows, environments, variables and datasets, built entirely from the conversation.
create_api_collectionadd_api_endpointcreate_api_flowcreate_api_datasetimport_api_wsdland moreAssert and verify APIs
Assertions, contract schemas, mock servers and webhook inboxes, so an agent can prove an API behaves rather than just call it.
add_api_assertionsave_api_schemacreate_api_mock_servercreate_api_webhook_inboxrun_api_flowand moreReach private APIs
Client certificates, CA bundles, auth profiles and local agents, for services that never leave your network.
create_api_client_certificatecreate_api_ca_bundlecreate_api_local_agentcreate_api_auth_profilerevoke_api_tokenand moreWork with your tracker
Pull requirements in from Jira, Azure DevOps, TestRail or Zephyr, and push comments, transitions and defects back out.
import_jiratransition_jira_issuefile_jira_defectraise_bugimport_testrailand moreAsk about quality
Answer the questions people actually ask: what is covered, what is failing, and whether this release is safe.
get_coverageget_project_analyticslist_quality_gatesand moreSchedules and settings
Put runs on a cadence, pair a local agent for private environments, and manage workspace settings.
create_scheduleset_schedule_enabledpair_agentupdate_settingsstart_tourand moreCloud browsers
Run against a wider browser and device matrix through a connected BrowserStack account.
run_tests_browserstackget_browserstack_configsand moreConnect it
Two minutes, one API key
Generate a key in Settings, drop the config into your client, restart it. That is the whole setup.
One command, then restart your session.
claude mcp add axonqa \
--transport http \
https://axonqa.com/api/mcp \
--header "Authorization: Bearer YOUR_API_KEY"Replace YOUR_API_KEY with a key from Settings, API Keys. Keys can be given an expiry and revoked at any time.
Safety
An agent with real permissions needs real limits
Everything an agent does through MCP is authenticated, bounded, and written down.
Keys you control
Bearer tokens you generate, scope to an expiry date, and revoke the moment you want to.
Confirmation to destroy
Destructive and selected high-impact actions ask first. An agent cannot quietly remove your work.
Every call logged
Each tool call is recorded in the audit trail, so you can always see what ran and when.
Reference
POST /api/mcp | JSON-RPC endpoint. Supports initialize, tools/list, tools/call, tools/info and ping. |
GET /api/mcp | The live tool list, always current. |
/api/mcp/stream | Server-sent events for long-running tools such as crawls and runs. |
/api/mcp/openapi.json | OpenAPI 3.1 spec, for agents that prefer function calling. |
Rate limit | 100 requests a minute, with X-RateLimit headers and Retry-After on 429. |
Prefer to read the live list? Fetch the tools straight from the server
Point your agent at your own app.
Choose a plan, generate an API key, and let your assistant run its first suite in minutes.